Acta Montanistica Slovaca

Ronk 10 (2005), mimoriadne slo 1, 121-126

Experience in coal sampling parameters calculation

Tomanec Rudolf1 and Milovanovic Dejan
Sksenosti s vpotom parametrov vzorkovania uhlia
This paper describes the selection of parameters neccessary for the calculation of minimal representative quantity of coal sample
for various investigations in the field of mineral processing. The procedure is illustrated on the example case of the Drmno coal field,
from where the coal is delivered to the thermal power plant Kostolac. Two primary samples of coal, crushed down to different sizes and
prepared (homogenized and divided), are analyzed for the ash content and the results were statisticaly processed. The analytical results
are statistically processed. Analytical and graphical solutions, at the lowest sampling error, give an optimum range of the representative
minimal sample mass for the given coal size. The coefficient of proportionality, k, which characterizes the kind of mineralization in the
given material, and exponent for the Drmno coal type, are determined.
Key words: primary sampling, coal, representative secondary sample, statistical errors

Introduction
Klenovik, irikovac, and Drmno are three open-pit mines of the Kostolac coal basin, which produce
significant quantities of coal. The run-of-mine (ROM) coal is crushed and screened prior to its conveying
to the Kostolac B thermal power plant (TPP). The steam coal for the TPP is permanently sampled to control
the coal quality for coal quality control. Both, the coal producer and the user (TPP) are equally interested
in the operative-control sampling and the quality control due to technological demands and mutual
commitments.
In the coal preparation plant a mechanical sampling system, was used designed to perform the sampling
and the sample partition, on the spot, with a minimal representative mass for the required analyses.
The sampling system was planed to operate according to the valid standards and some empirical knowledge
of coal sampling of similar properties. Unfortunately, though needed, the device has not been put into
operation.
The aim of this work is to identify the sampling parameters for the Kostolac coal type mined
at the Drmno open pit, providing the sampling principles, which determine the minimal quantity
of the representative sample for the given raw material properties. For the experimental selection
of parameters and the statistical data processing, two samples of different size and weight were taken and
used in this investigation.
Sampling
According to the designed and partly completed technology, the coal preparation process consists
of the following:

separation of lump coal (for domestic purposes) on roller rosts and a further classification,

crushing of the undersize in hammer crushers down to -30 mm (for TPP) and conveying to the TPP
stockyard,

alternatively, the ROM coal can be routed to crushing for the steam coal production, which is usualy
perfomed during a winter season, when fuel is increasingly demanded by the TPP.
After crushing, the grain class of steam coal -30+0 mm, which is the subject of our test, is conveyed
by the belt system to the TPP
The task of the test was taking the sample from the crushed coal stream, size -30+0 mm,
and its preparation with the purpose of establishing the minimal sample quantity for the given analytical error
for the specific type of coal.
A experimental of sampling parameters for this coal a significant mass of the primary sample.
The principal assumption of the sampling theory is based on the fact that the sampling error is inversely
related to the sample mass, which means that an increase in the sample mass will provide a lower sampling
error.
For the primary sampling, the well-known method [1-6] was used, which basically assumes the relation
1

prof. Dr. Tomanec Rudolf, prof. Dr. Milovanovic Dejan, Faculty of Mining and Geology, University of Belgrade, 7. Djusina Str.,
11000 Belgrade, Serbia, tomanec@rgf.bg.ac.yu
(Recenzovan a revidovan verzia dodan 21. 10. 2005)

121

Tomanec Rudolf and Milovanovic Dejan: Experience in coal sampling parameters calculation

Qprim = m n ,
(1)
meaning primary sample mass is proportional to the minimal mass of individual samples (increments) m (kg)
and to the number of individual samples required n. The minimal mass of individual samples will depend,
in the given case, on the conveyor-belt capacity (Q = 2700 t/hr), its width (b = 1600 mm), moving speed
(v = 3,9 m/s), and the total length (L = 30 m). The sampling interval is calculated from the sampling
consignment (M), the conveyor-belt capacity (Q), and the empirically determined number of individual
samples. The mass of the individual sample can be expressed as:
Q b sin
(2)
3.6 v
where, b is the slot width of the sampler (mm), is the angle between the conveyor centreline
and the sampler and v is the conveyor velocity (m/s). The sampling interval for the individual sample
separation from the conveyor will be:
m=

60 M
(3)
Qn
where, M is the consignment to be sampled (t), Q is the belt capacity (t/hr), and n is the number
of the individual samples.
In the case of sampling by the mechanical sampling system, from the standing or moving conveyor belt,
the sample mass (kg/hr) can be calculated from the formula:
ti =

Q b n meh
(4)
3.6 v T
where T is the sampling time (hr), n is the number of partial samples per hour, and b is the width
of the sampler (b > 3d).
Certainly, for a manual sampling, when the belt stands still, the calculation takes into consideration
the length (L) of the conveyor belt on which the material is sampled. The samples for this investigation were
taken from the belt conveyer TR - 10, where the mechanical sampling system is located according
to the design, and prepared for the laboratory tests. For the reported test, two representative samples of raw
coal were separated.
The first primary sample 1Qprim, size -40+0 mm, was collected from the conveyor by the partial cutting
method. The material was cut on the conveyor belt in a length of 2 m. Individual samples were unified into
composite samples within the time interval of one hour. The Sampling proceeded in the same way, each hour
through one shift, and the finally collected material formed one composite sample. The total mass of this
representative sample was 1200 kg. The second sample 2Qprim, was simultaneously collected in the same
manner and weighted 600 kg. Prior to the dividing, this sample was crushed down to the size of -10+0 mm
in the jaw crusher.
Both representative primary samples: first one 1Qprim of 1200 kg weight, size -40+0 mm, top size
37,4 mm; and the second one 2Qprim of 600 kg, size -10+0 mm, top size 5,83 mm, were equally homogenised
and divided into eight individual samples in six sets of secondary samples. The treatment procedure
of the two representative raw coal samples is shown in Fig. 1. In each of the six stages of the preparation
(homogenisation, quartering, homogenisation, chessboard sampling), the primary sample was divided
and eight representative samples were obtained for chemical analysis. After the treatment, six sets of eight
single representative samples were separated. Each sample in a set was analysed for the ash content,
statistically processed, and average ash content p i , the average deviation p, ash variation coefficient
KV were expressed for each set. Finally, the mean error of the ash content determination or the laboratory
analytical error was determined. The respective mean values are given for each of the six sets of the samples
in tables for 1Qprim on the left and 2Qprim on the right of the diagram (Fig. 1 ).
The average ash content for the first sample 1Qprim was 24,74 %, which was used to calculate
the deviation from the average rate for each set of samples. It proved, as mentioned in the beginning, that
with the increasing masses of individual samples from 2,34 to about 75 kg, the mean determination error
of ash content in samples decreases from 2,55 to 1,62 %, which was extremely helpful in the analysis
of the representativiness of the minimal sample mass. Ranges of other calculated parameters, from the set
I to VI, were from 1,83 to 3,35 for the standard deviation, and from 7,59 to 14,71 % for ash content variation
coefficient.
The average ash rate for the second sample 2Qprim was 24,75 %. Like for the first sample, the deviation
from the average content was calculated for each set. With the increasing mass of individual samples from
1,17 to 37,5 kg, the mean deviation decreased from 2,67 to 1,68. Ranges of other calculated parameters, from
Qmeh =

122

Acta Montanistica Slovaca

Ronk 10 (2005), mimoriadne slo 1, 121-126

set I to VI for the second sample, were from 2,12 to 3,14 for the standard deviation and from 8,05 to 12,03 %
for the ash content variation coefficient. As before, the highest deviation from the standard ash content was in
the set VI (inadequate sample weights).

Fig. l. Summary of sampling stages employed in this examination.

The relationships of individual sample weights and errors, (deviations from the mathematical
expectations), for both tested samples are presented graphically in Fig. 2. Beside curves 1 and 2,
the analytical error of ash content determination is marked as the minimal value - sensitivity step
of the specific laboratory.

123

Tomanec Rudolf and Milovanovic Dejan: Experience in coal sampling parameters calculation

Fig. 2. Sample weight versus the error

of the determination of representative sample
minimal mass.

3.0
2.8
2.6
2.4
2.2
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0

B
A

10

20

30

40

50

60

70

80

90

100

Weight, kg

Selection of sampling parameters

The two obtained curves can be approximated by the analytical relation y=a/x+b. The curves show that
with the increasing mass of individual sample, the mean deviation gradually decreases to a constant value,
(irrespective of the further increase of the sample mass) which represents the analytical error or sensitivity
threshold for the applied method. The absolute value of deviation cannot be reduced under the given
conditions below this minimum (rectilinear part of the curve). Since the rectilinear parts of the curves, do not
concurr, their mean value is drawn as a line between them.
The mean value of 1,65 is the analytical error, unavoidable for the given technique of the sample
treatment and the chemical analysis in the given laboratory. The minimal masses of representative samples
are determined on this intermediate line. The minimal mass of the representative sample 1Qmin is obtained
as follows: A vertical line is drawn from the point A (the intersection of the line a - the mean deviation from
the average rate and the curve 2) up to the curve 1. In the point B (intersection of the vertical line
and the curve 1), the tangent t is set intersecting the line a and the abscissa wt. The minimal mass
of representative sample 2Qmin is obtained in a similar way: the vertical line is drawn from the intersecting
point A of the line a and the curve 2 to the abscissa where the value of the minimal mass is read for
the representative sample Q2 16,70 kg. In this way, the applied statistical data processing and the graphical
method provided the minimal mass of the representative samples adequate to the given conditions
and the known analytical error.
These values, were used for the determination of basic parameters in the equation Qmin = kdmax.
The first primary sample 1Qprim = 1200 kg
Size fraction -40 + 0 mm
pi = 24,74 %
mi = 2,34 74,5 kg
The mean determination error
p = 2,5 1,62 %
p =1,83 3,35
KV = 7,59 14,71 %

The second primary sample 2Qprim = 600 kg

Size fraction -10+ 0 mm
p i = 24,75 %
m1 = 1,17 37,5 kg
The mean determination error
p = 2,67 1,68 %
p = 2,12 3,14
KV = 8,05 12,03 %

The values obtained for the minimal mass of the samples 1Qmin and 2Qmin were substituted in the above
equation to determine the parameter k and as follows: the established graphical representation
of the relationship between the mean error and the sample mass Q gives the respective masses of samples
Q1 and Q2, and the mean deviation of 1,65 % for samples crushed to sizes dl and d2. By substituting
the respective Q1 and Q2 masses in the equations:
Q1 = k d1
(5)
Q2 = k d 2

and taking the logarithm of these equations, the results will be:
log Q1 = log k + log d1

124

(6)

(7)

Acta Montanistica Slovaca

log Q2 = log k + log d2

Ronk 10 (2005), mimoriadne slo 1, 121-126

(8)

The subtraction of both equations is:

log Q1 log Q2 = (log d1 log d2)
Q
log 1
log Q1 - log Q 2
Q2
=
=
and
d
log d 1 - log d 2
log 1
d2
Substituting in the first equation, will give the value of k:
log k = log Q1 log d1

(9)
(10)

(11)

According to the published papers and references [1,7], the value commonly used for the coefficient
k is 2 or higher. As the determination of sampling parameters requires coefficient k, which is directly
dependent on the type and character of material, the provisionally accepted value of 2 and its substitution
in the above equation will give 0,07.
The overview of the well-known and mostly used principles for determination of the minimal sample
mass for various technological investigations is given in Tab. l.

1.

2.

3.

Tab. 1. Overview of methods for the determination of the minimal mass of representative samples and respective parameters
Referencee
Equation
Parameters
Richards 1909;
Qmin - minimal quantity of sample, kg; dmax top size, mm; k
2
Qmin = k d max
G. O. Chechott
coefficient of proportionality which depends on material
1920; 1932
characteristics: variation in the valuable component content,
mineral size, and the component content in the raw material (0,1624,00, presently in practice 0,1-0,5-1,0). For coal (0,4-1,5).
Demond and
Qmin - minimal sample mass, kg; dmax top size, mm; k

Qmin = k d max
Halferdal 1923.
coefficient of proportionality, which depends on material
properties (density, grain shape, content, uniformity and size of
minerals) and on the permissible error; - exponent value which
varies in relation to the raw material within the range from 1,5 to
2,7. Coefficients k and are empirical given in a table.
Vezin 1865.
Q - initial sample mass (primary sample); q - sample weighted for
3
q = K d
the chemical analysis (secondary sample); d grain top size in the
sample for the chemical analysis; D - grain top size in the primary
q d 3 n d 3
= 3
= 3
sample; raw material density; n - number of mineral grains.

D n

4.

Brunton 1898.

5.

Chernenko.

6.

P. Gy 1982.

7.

N. Barishev

8.

I.P. Alimarin

Q=

D f n

1
10 4 p

Q = k d
Qmin =

Pozharickii

C d3
V

Q = qn = q

t 2 Vm 2
p2

(100 P ) d 3

Q
9.

n 2 P 10 5

Qmin = 0,06 d 2
Qmin = 0,1 d 2
Qmin = 0,18 d

10.

G.I.Prejgerzon,
M.N.Albov,
V.M.Kryter

2.25

Q = qn

kg
kg

Q - primary sample minimal mass, kg; D top size of the grains in

the sample, mm; - valuable mineral density; - metal content in
the valuable mineral; - average metal content in the sample; n - number (quantity) of grains in the divided sample suffocate or
deficit regarding the primary sample; f - rate of largest valuable
mineral grains; p - permissible sampling error, equal to the relative
error of the chemical analysis.
Q - representative sample minimal mass; d - mean diameter of
largest grains, mm; k - coefficient of proportionality, 1-1,5 for the
uniform and non-uniform mineralisation.
Qmin - minimal permissible sample mass for the analysis, g; C raw material constant, g/cm3; d top size of sampled material, cm;
V - variation coefficient of valuable mineral content.
q mean mass of single samples, kg; n - number of single samples
t - probability coefficient equivalent to the sample
representativeness (t<1,7-2,0); Vm - variation coefficient of
valuable component in the sample, p - given sampling error
P - valuable component content, %; d top size, mm;
- average particle density; n permissible relative analytical
error
- Uniform mineralisation, valuable component with small
intergrown grains
- Non-uniform mineralisation, like rare mineral ores.
- Extremely non-uniform distribution of mineral component.

kg
q mean mass of individual samples (portions, cuts) to be unified
into a general sample, kg; n number of individual samples,
where n = (tV/p)2; V - coefficient of metal content variation in
single samples, %, t - probability coefficient; po - permissible
(given) sampling error, % of the component content

125

Tomanec Rudolf and Milovanovic Dejan: Experience in coal sampling parameters calculation

The method determining k and is based on the assumption that with the increasing sample mass
(for given size of material) the sampling error decreases to a value which depends on the accuracy
of chemical analysis and the sample preparation method. Thus, the summary or the total error of sample
preparation, excluding the sampling error, consists of the error of the sample division due to the inadequate
homogenisation during the treatment and error from insufficient number of pieces in the sample, i.e.
the inadequate sample mass, as well as the error of the chemical analysis [8]. Since the error from
the insufficient sample mass is included in the total error, and because the error resulting from the inaccuracy
of the sample division is insuperable, the former error is not possible to determine. Still, a careful operation
of the sample division during the homogenisation can reduce the specific error to a constant value.
Consequently, the variation of the total error from one set to another will be a result of the disproportion
between the sample mass and the grain size in the mixture.

Conclusions
The described method of the determination of the parameter , and even more of proportionality
coefficient k, is a quite complex and tedious procedure due to a large number of samples to be treated, which
is time and money consuming. In turn, it provides parameters whose values are accurately determined
characteristics of the given coal type, applicable to the given deposit for a long time (uniformity
and constancy of mineralisation).
A primary sample, which would be separated in the designed sampling system at the transfer point
of conveyor belt T-10, would weigh about 1730 kg (single samples of 57 kg, sampling interval of 16 min).
The mass of the representative sample of the given coal size, hauled to the TPP stockyard, is calculated,
2
on the principles which assume that Q min = k d max
, to be k =0,10 and =2, and amounts 160 kg.
The preparation of the primary sample, specially separated for this investigations, with the total mass
of 1800 kg (1Qmin =1200 kg and 2Qmin =600 kg), the statistical processing, i.e. the calculation of the deviation
from mathematical expectation at the given level of data reliability (analytical error), and the application
of the graphical method, all gave the representative sample minimal mass of 16,70 kg for the finer coal
or 92,75 kg for the steam coal, with the size of-40+0 mm.

References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]

Albov, M., N.: Oprobovanie mestorozhdenii poleznykh iskopaemykh, Nedra,1975, Moskva.

Artiushin, S., P.: Sbornik zadach po obogashcheniiu uglei. Nedra, 1979, Moskva.
Gy, P., M.: Sampling of Particulate Materials., Elsevier, 1982, Amsterdam.
Kallistov, P., L.: Uchet vysokikh prob i samorodkov pri podschete zapasov mestorozhdenii zolota;
Glavspetstsvetmeta, 1952, Moskva.
Khan, G., A.: Oprobovanie i kontrol tekhnologicheskikh protsessov obogashcheniia, Nedra, 1979,
Moskva.
Ogloblin, N., D., Mrushlevich, I., V.: Tehnicheskii kontrol na ugleobogatitelnykh fabrikakh,
Ugletekhizdat, 1958, Moskva.
Sagradian, A.L., et al.: Kontrol tekhnologicheskogo protsessa flotatsionnikh fabrik; Nedra, 1983,
Moskva.
Sergo, E., E.: Oprobovanie i kontrol tekhnologicheskikh protsessov na obogatitelnykh fabrikakh,
Vishcha shkola, 1974, Kiev.
Tomanec, R., et al.: Coal sampling parameters for Drmno coal field, Kostolac - Proceedings of 8th
Balkan conference on Mineral Processing, Volume one, p. 11-18. Belgrade, 1999, Yugoslavia.

126